NextFin News - Europe is struggling to get space-based data centers off the ground. While American startups have already launched hardware carrying data-center-class GPUs into orbit and China has deployed the first operational space-computing constellation, Europe is still weighing whether to build orbital data centers at all - a delay that a European think tank warns could leave the continent dependent on foreign orbital computing infrastructure.
While European actors continue to weigh the development of orbital data centers (ODCs), Chinese institutions and companies are already moving towards implementation.
The European Space Policy Institute said in a report published Aug. 5 that the gap between European deliberation and Chinese execution is widening. The warning reframes a technology that began as a research curiosity into a sovereignty contest with the same shape as the semiconductor race: the United States is moving through private capital, China through state coordination, and Europe through consultations.
The Race Has Already Moved Past the Feasibility Debate
The premise is simple. Place data centers in low-Earth orbit, where sunlight is uninterrupted, waste heat radiates into the vacuum, and there is no grid connection to wait a decade for. The idea has grown urgent because terrestrial AI data centers are running into hard constraints: electricity connections, water for cooling, land use, and local opposition. A European Parliament research service paper published in 2026 concluded that the main barrier to space-based data centers is economic rather than technical, with future launch costs the key determinant - but also that the concept raises unresolved questions over data and orbital regulation that may require updating existing digital and space treaties.
The problem for Europe is that the rest of the world has stopped debating and started launching.
In the United States, Starcloud - formerly Lumen Orbit, a Y Combinator-backed company founded in January 2024 - launched its first satellite in November 2025, just 21 months after founding. The 60-kilogram Starcloud-1, about the size of a small refrigerator, carries an Nvidia H100 GPU that the company says is 100 times more powerful than any GPU previously operated in space. In December 2025 it trained a GPT-style model (Andrej Karpathy's NanoGPT, fed the complete works of Shakespeare) in orbit and ran inference on Google's Gemma - the first time a model of that architecture has been trained above the atmosphere. In March 2026 the company raised a $170 million Series A at a $1.1 billion valuation, co-led by Benchmark and EQT Ventures, bringing total capital raised to $200 million and making it the fastest unicorn in Y Combinator history. Its second satellite, with 100 times the power generation of the first and carrying multiple H100s plus Nvidia's Blackwell B200 chip, is scheduled for October 2026.
Axiom Space, the Houston-based space-station developer, deployed a data-processing prototype called AxDCU-1 on the International Space Station in the fall of 2025 and launched its first two orbital data center nodes to low-Earth orbit on Jan. 11, 2026, operating on Kepler Communications' optical relay network.
China has moved faster still. ADA Space, working with the Zhejiang Lab, launched 12 satellites on May 14, 2025, aboard a Long March 2D rocket from Jiuquan - the first tranche of a proposed 2,800-satellite "Three-Body Computing Constellation" with a combined 5 peta-operations-per-second of compute and 30 terabytes of onboard storage. By February 2026, Beijing said in-orbit tests had demonstrated core capabilities including networking, computing, model deployment and scientific payload verification. In July 2026, Shanghai Xingshu Tiansuan Space Technology launched the first constellation of a planned 1,000-satellite space-computing network. Other Chinese players include China Mobile, Comospace, Bailing Aerospace and Orbital Chenguang, and in June 2026 municipal science and technology agencies in Beijing issued a call for proposals on enabling technologies - part of what the think tank describes as a state-coordinated industrial framework.
Against that backdrop, Europe's position looks hesitant. The continent has supported several technologies relevant to orbital computing - the European Space Agency's Φsat series, the ReOrbit satellite working with ESA's InCubed program to demonstrate secure space-to-space and space-to-ground data transfer, and research into space-based data centers. But the think tank found that Europe has yet to organize these efforts around a large-scale deployment strategy. Commercial projects, including plans announced earlier this year by French startup ALATYR for robotically assembled orbital data centers with megawatt-class capacity, have not matched the scale seen in the United States and China. The European Commission held a workshop on orbital data centers in Brussels on May 13, 2026, gathering end-users, industry and researchers. The machinery is still in the consultation phase while hardware is already in orbit.
Why Europe Is Behind: Three Structural Disadvantages
The gap is not an accident of corporate strategy. It reflects three structural differences between Europe and its competitors.
First, capital structure. The American model is private risk capital at speed. Starcloud went from founding to a $1.1 billion valuation in roughly two years. That model tolerates the kind of failure a public program cannot: a single launch failure is a write-down, not a political scandal. China's model sits at the opposite pole: state credit, municipal initiatives and state-owned enterprise backing. The think tank reported that Chinese orbital-computing ventures have been backed by billions in credit lines, and that Beijing views space-based compute as the next frontier of its broader technology strategy - a way to produce advanced computing domestically, export it, and position itself to set global standards and regulation before others get there first.
Europe has neither. Its space industry is dominated by prime contractors answering to multi-year government procurement cycles, while its venture ecosystem is smaller and more risk-averse than Silicon Valley's. A European startup with a Starcloud-style timeline would struggle to find a funding path that does not run through Brussels - and by the time it does, the technology has moved.
Second, the demand anchor. Both the United States and China have a clear first customer for orbital computing. In America, it is the commercial satellite and AI sector: Starcloud's near-term business is processing Earth-observation data in orbit so that only insights, not raw imagery, need to be downlinked. Axiom's nodes are aimed at government and commercial customers needing secure, cloud-enabled processing for satellites and spacecraft. In China, it is the state itself - a coordinated buyer with sovereignty objectives. Europe has large Earth-observation assets - the Copernicus constellation is among the world's most capable - but it has not signaled demand. The think tank's central recommendation is precisely that the ESA and the EU act as early customers by offering contracts for in-space data processing services. The absence of that signal is the absence of a market.
Third, regulatory coherence. Orbital data centers sit at the intersection of three regimes that Europe has not harmonized: spectrum allocation through the International Telecommunication Union, national launch and operating licenses, and emerging space-traffic management rules. The European Parliament's research service noted that space data centers may require updating or clarifying existing digital and space treaties and legislation. Fragmented regulation is a tax on any new industry; for one that operates across borders by definition, it is a barrier to entry.
The Technology Is Real - but the Economics Are Not Yet Proven
The strongest argument against Europe rushing in is that the business case is unproven. The parliamentary paper is explicit: the main barrier is economic, not technical. Launch costs are the key determinant, and while reusable rockets have cut the price of reaching orbit dramatically, a data center is not a single satellite. It is a power plant, a cooling system, a computer and a communications array - and every watt of compute requires watts of solar generation, storage for eclipse periods, and bandwidth to move data to and from the ground.
There are harder technical problems too. The radiation environment degrades electronics, and industry analyses note that radiation-hardened AI accelerators lag their terrestrial counterparts by two to three technology generations - meaning an orbital data center risks running yesterday's chips at tomorrow's cost. Tens of thousands of tracked objects crowd low-Earth orbit, creating collision risk and insurance costs. Cybersecurity is a novel attack surface: satellite command-and-control links and inter-satellite networking are not covered by terrestrial security frameworks.
The think tank itself tempers the hype: more ambitious uses, such as globally distributed computing networks and large-scale AI training envisioned by companies including SpaceX, remain more than a decade away. The first commercially viable applications are likely to involve processing Earth-observation data close to where it is collected - a narrower, more plausible market than "AI training in space."
That is a legitimate reason for caution. But caution is not the same as strategy. Europe's error is not that it has declined to spend billions on unproven infrastructure. It is that it has not made the smaller, cheaper decisions that would let a market form: acting as an anchor customer, coordinating spectrum and licensing, and funding the demonstrations that de-risk private capital.
The Second-Order Risk: Dependence, Not Just Lost Revenue
The obvious cost of falling behind is lost industrial revenue - European companies missing out on a new market. The deeper risk is dependence. If orbital computing becomes critical infrastructure - for real-time Earth observation, for secure government communications, for AI inference at the edge - then whoever owns the orbital layer controls the terms of access. The think tank's warning is framed in exactly these terms: Europe risks "relinquishing space compute capabilities to the United States and China."
This is the same logic that drove Europe's push for semiconductor sovereignty after the chip shortage. The difference is that chips can be fabricated on European soil. Orbital infrastructure, once launched, cannot be recalled or relocated. A European satellite that depends on a foreign orbital data center for processing has no fallback.
There is also a standards dimension. Beijing is explicitly positioning itself to help set the global standards and regulation for the industry before others arrive. Standards in space - spectrum coordination, docking interfaces, data formats, debris-mitigation practices - tend to stick once established, because the cost of switching an orbital architecture is prohibitive. Europe's traditional strength is writing rules; its weakness is that it is writing them for a technology that others are already deploying.
The Counter-Thesis: Europe May Be Right to Wait
The strongest case against the alarm is that the technology may not work at scale - and that Europe, by waiting, avoids wasting public money on a failed bet.
The economics are genuinely uncertain. Even at current launch prices, lifting the mass required for a meaningful data center - solar arrays, radiators, compute, shielding - is expensive, and the revenue model rests on assumptions about downlink bandwidth savings and latency sensitivity that have not been proven in a commercial market. The radiation problem means hardware must be replaced more often than on Earth, adding to lifecycle costs. If terrestrial solutions - new nuclear reactors, grid upgrades, more efficient chips - solve the power bottleneck faster than expected, the space value proposition collapses.
There is also a strategic argument for letting the private sector lead. If orbital data centers prove viable, European companies can enter later as customers or niche suppliers without having pre-committed public capital. Europe already has strengths it can leverage: the ESA's technical capability, the Copernicus and Galileo constellations as potential anchor workloads, and a regulatory apparatus that could shape global standards if it moves decisively.
This case is coherent. But it depends on one assumption: that Europe retains the option to enter later on favorable terms. If the orbital layer consolidates around American and Chinese providers - with proprietary interfaces, exclusive spectrum arrangements, and security restrictions - that option may not exist. Waiting is a strategy only if the door stays open. The evidence so far suggests it is closing.
What to Watch: The Next Two Years Decide It
The verdict on Europe's orbital data center effort is not that it has failed technically. It is that it has failed to decide. The technology has moved from concept to demonstration - Starcloud trained a model in orbit; Axiom has nodes in space; China has a constellation operating. The question is no longer whether orbital computing works. It is who will own it.
Short term (1-3 years): Expect continued fragmentation. European startups will announce partnerships and demonstrations; the EU will hold workshops and issue studies; no flagship program will launch. The market will be defined by the first commercially viable use case - processing Earth-observation data in orbit - and European Earth-observation operators will face a choice between American and Chinese processing infrastructure.
Medium term (3-7 years): The divergence widens. If the EU follows the think tank's recommendation and establishes a flagship ODC initiative under its 2028-2034 research program, with the ESA and EU acting as anchor customers, European companies could capture a sovereign niche. If it does not, European demand will be served by foreign providers, and Europe's role will be limited to component supply.
Long term (7+ years): The orbital layer becomes infrastructure. If large-scale AI training moves to orbit as companies including SpaceX envision, the owners of that capacity will hold a structural advantage. Europe's position then depends on decisions made in the next two years.
The falsifying signal is specific: if the EU commits to a flagship orbital data center initiative with anchor customer contracts by the time its 2028-2034 research framework is finalized - and European startups secure launch commitments on a timeline comparable to Starcloud's - the "Europe is falling behind" thesis is wrong. If the 2028-2034 framework passes without a dedicated ODC line and European demand continues to be served by foreign infrastructure, the thesis is confirmed.
The beneficiaries of inaction are clear: American venture-backed startups and Chinese state-backed constellations. The exposed are European satellite operators, defense contractors, and any industry that depends on sovereign, low-latency data processing.
Here is the uncomfortable truth for European policymakers: the orbital data center race will not be won by the continent that writes the best white paper. It will be won by the one that launches first, buys first, and accepts that the price of sovereignty is paid in capital, not in consultations.
Explore more exclusive insights at nextfin.ai.
